You are on page 1of 8

Proceedings of the ASME 29th International Conference on Offshore Mechanics and Arctic Engineering OMAE2010 June 6-11, 2010,

Shanghai, China

OMAE2010-20520
Vortex Induced Vibration Analysis of a Complex Subsea Jumper
Samuel Holmes Red Wing Engineering, Inc. Yiannis Constantinides Chevron Energy Technology Company

ABSTRACT Jumpers are typically short sections of curved pipe spanning production riser elements on the sea floor. When in areas of significant currents these jumpers are subject to vortex induced vibration (VIV). The complex shape of the jumper means that numerical methods are usually needed to solve for the vibration modes of the jumper. Furthermore, the fluid flow around the jumper is also complex so that traditional methods of VIV analysis used for risers are not applicable to jumpers. Here we use a CFD code in a fully coupled analysis to predict vibration response and strain of a typical subsea jumper. A separate finite element analysis is used to calculate the eigenvalues and eigenvectors of the jumper system for input into the CFD analysis. The resulting method is economical and practical for design analyses. INTRODUCTION The assessment of the fatigue life of a subsea jumper subject to VIV is a difficult and important engineering problem. The remoteness of the jumper makes inspection and servicing difficult while the complex shape of the jumper means that accurate estimation of vibration amplitudes is not possible with traditional analysis methods. In this work we show a relatively simple and straightforward analysis that predicts jumper response. The objective of this work is to assess the potential vortex induced vibration (VIV) of a subsea jumper line subjected to a steady current. VIV may cause vibration leading to high cycle fatigue of the jumper which is also subject to stress corrosion effects. In this case, the subsea jumper transfers water between two points about 30m apart on the seabed. The jumper is made from thick welded steel pipe jumper has an irregular shape as shown in blue in Figure 1. As designed, the jumper also has a center buoyancy section (Figure 1, green section) made of syntactic foam to provide lift at the unsupported center of the jumper.

This center section is straked to reduce vortex induced vibration. In this example we studied the case in which the current direction is normal to the jumper as shown in Figure 1 and current speeds range from 0.1m/s to 0.5m/s.

Current

Figure 1 Jumper Geometry This analysis method used here combines a linear structural vibration model (modal analysis) with a computational fluid dynamics (CFD) model of the fluid. The structural model is based on an eigenvalue analysis from the Abaqus1 structural analysis code. We expect the jumper motions to be relatively small and the jumper material to remain elastic so a linear material model is used and small deflections are assumed. This external analysis is used to derive the eigenvectors (mode shapes), frequencies and masses required for the analysis. The complete fluid-structure interaction solution is then solved in the CFD solver which performs the model vibration analysis as part of the solution.
1 Abaqus is a registered trademark of Dassault Systmes or its subsidiaries in the United States and/or other countries

Copyright 2010 by ASME

APPROACH Because we wanted to focus as quickly as possible on potential VIV problems we chose to model cases that seemed most likely to cause VIV. Thus we chose to have the current flow direction normal to the plane of the jumper and chose current speeds that gave vortex shedding frequencies at or near expected natural response frequencies. Two basic jumper configurations were tested. The first is the baseline or installed design which includes the floatation section and its strakes. The second configuration is a plane bare jumper with no floatation or strakes. Structural Modeling - The jumper pipe is a short heavy structure with a density about 3 times its displaced volume. To make the jumper, seven short sections of straight pipe are welded together to form a serpentine pipe lying in a plane as shown in Figure 2. The jumper spans slightly more than 30m and the lowermost horizontal section is about 3m from the sea floor. The cross section and material properties used in the structural analysis of the jumper are given in Table 1.

Table 1 Jumper section dimensions and material properties OD (m) Jumper Buoyancy Section 0.167 0.6 Material Density (kg/m3) 7849 509

Steel Syntactic Foam

Table 2 Jumper natural frequencies in air Mode Number 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 Frequency (rad/s) 2.2977 4.3371 6.1839 6.1897 9.7613 10.584 15.151 17.569 22.802 23.545 29.385 31.352 32.109 59.157 59.318 76.805 79.097 84.010 94.326 99.090 Frequency (Hz) 0.36569 0.69026 0.98420 0.98511 1.5536 1.6845 2.4114 2.7962 3.6290 3.7473 4.6768 4.9899 5.1103 9.4152 9.4408 12.224 12.589 13.371 15.012 15.771 Generalized Mass (kg) 250.94 262.62 170.25 389.21 61.667 279.20 85.032 123.26 29.714 109.40 38.947 167.38 104.38 112.55 114.17 57.233 44.721 72.998 97.297 111.27

Figure 2 Jumper Dimensions

The structural analysis characterized riser response using a total of a total of 20 modes. It should be noted that the modal response analysis characterizes the structural response without including the effects of the surrounding fluid, i.e. without any added mass effects. These effects are automatically incorporated when the fluid and solid solutions are combined by the flow solver. Table 2 lists the mode numbers frequencies and associated masses from the eigenvalue analysis. Note that the frequencies in air range from 0.36 Hz to 15 Hz. We estimated the effect of the added mass of the surrounding water using an added mass coefficient of 1 and found that the added mass reduces the frequency range to 0.27 Hz to 13 Hz. The maximum primary vortex shedding frequency of a 0.167m diameter pipe at as current speed of 0.5 m/s is 0.6 Hz and the most significant exciting frequencies do not exceed three times the primary frequency (or about 2.4 Hz in this case) so we do not expect to excite the higher modes significantly. Thus modes above mode 8 are not expected to be significant in this flow environment. The mode shapes associated with the modes vary quite a bit from mode to mode. The first four mode shapes are shown in Figures 3 to 6.

Static Mode 1 20 15 y(ft) 10 5 0 100 50 x(ft) 0 -0.5 -1 z(ft) 0.5 0 1

Figure 3 Mode shape for Mode 1

Copyright 2010 by ASME

Static Mode 2 20 15 10 5 0 100 50 0 -0.5 -1 0.5 0 1

Fluid flow analysis The fluid flow used the finite element solver AcuSolveTM. This solver is second order accurate in both space and time and uses unstructured grids and supports a variety of element types. In the meshes shown here, 6 node wedge elements are used in the boundary layer and, 4 node tetrahedral elements are used outside the boundary layer. AcusolveTM also supports a variety of turbulence models including Spalart-Allmaras (S-A) model [1], the dynamic subgrid model, Large Eddy Simulation (LES) and Detached Eddy Simulation (DES) based on the S-A model [2]. All of the simulations reported here used the DES turbulence model. We chose a large flow domain around the jumper to model the fluid flow as shown in Figure 7. This domain is a rectangular block with a width of 33.48m and height of 10m and a length in the flow direction of 10m. As shown in the figure, mesh has an extensive wake region of fine elements that extends to the sea floor behind the jumper. Other views of the mesh of the jumper with the buoyancy sections are shown in Figures 8 and 9.

Figure 4 Mode shape for Mode 2

Static Mode 3 20 15 y(ft) 10 5 0 100 50 x(ft) 0 -0.5 -1 z(ft) 0.5 0 1

FLOW

Figure 5 Mode shape for Mode 3

Static Mode 4 20 15 y(ft) 10 5 0 100 50 x(ft) 0 -0.5 -1 z(ft) 0.5 0 1

Figure 7 Overall fluid domain with jumper The mesh used for the jumper with the buoyancy sections contains a total of 4.1M nodes with 2.24M wedge elements in the boundary layer and 17.6M tet elements in the free stream. The boundary layer mesh was 7 elements thick with a first element thickness of 0.004m and a stretch of 1.2. The bare jumper mesh was similar to mesh of the jumper with the buoyancy section but was slightly larger with 4.5M nodes, 2.37M wedge elements and 19.3M tet elements. The mesh for the bare jumper was developed using the same techniques as those used for the first geometry. Figure 10 shows a view of this mesh.

Figure 6 Mode shape for Mode 4

Copyright 2010 by ASME

Sea Floor
Figure 8 Section at the middle of the jumper Figure 10 Jumper without buoyancy section flow analysis and so the modal analysis is easily carried out by the flow solver. However, it is necessary to calculate the mass and stiffness associated with each mode and the mode shapes beforehand so that these can be provided to the fluid solver. The application of this method to riser VIV has been described in more detail in References 3 and 4. Boundary conditions were used to move the nodes on the jumper surface and an arbitrary Eulerian-Langrangian (ALE) mesh movement algorithm was used to move the rest of the mesh. The problems generally ran at about 25 time steps per hour so the calculations took several days each. All of the calculations were started from established flow conditions, i.e. using previously calculate pressure, velocity and eddy viscosity nodal values in order to reduce computation time. RESULTS We first completed several calculations with the base configuration with the buoyancy section at current speeds of 0.3m/s and 0.5m/s. We tested the first case (0.3m/s) using 12 eigenmodes to characterize the structure and also using 6 eigenmodes to characterize the structure. We also varied the time step in the simulation. After some preliminary runs we settled on a time step of 0.02s at a current speed of 0.3m/s and a structural description using the first 6 eigenmodes because we saw little response associated with the higher mode numbers. Note that the expected primary vortex shedding frequency for the small diameter part of the jumper is 0.37Hz and the natural frequency of the 6th mode is 1.7Hz or about 4.5 times higher. These simple comparisons highlight an intrinsic advantage of using modal analysis to describe the structural response. By using only those modes whose frequencies are near the fluid excitation frequencies we can avoid numerical instabilities that might arise if a more complete structural model with a very

Figure 9 Surface mesh on buoyancy section Fluid-structure interaction - This fluid-structure interaction analysis requires simultaneous solution of the structural response and the fluid response. Here the structural analysis is performed completely within the CFD solver. The CFD solver solves for the fluid motions and the structural response in an iterative procedure within each time step. Satisfactory convergence between the two solutions usually took fewer than five iterations in the problems analyzed here. In general, the computational effort required to solve the structural vibration problem is a small fraction of the effort required for the fluid

Copyright 2010 by ASME

broad range of response frequencies is used. For example, problem can arise if the response frequencies of the model approach the Nyquist frequency of the time step [ f n = 1 /(2t ) ]. Besides the base configuration with the buoyancy section and strakes we built a second mesh for a configuration without the buoyancy section. This configuration was tested first at a current speed of 0.5 m/s and then at a current speed of 0.454 m/s. The second current speed was run to test for the effects of lock-in of the jumper VIV by aligning the vortex shedding frequency based on jumper diameter with one of the structures' natural frequencies (Mode 2 in this case using an estimated added mass). Case 1: Jumper with buoyancy section with current speed of 0.3 m/s The first simulation was completed at a current speed of 0.3m/s. We used 12 eigenmodes to characterize the jumper structural response. After some initial tests the final phase of the simulation was run for a total of 2,000 time steps simulating 40 seconds of physical time. Figure 11 shows the in-line displacement history of a point at the center of the buoyancy module. Note that in the mean displacement in the flow direction is just over 3 cm. The most striking result from the simulation was the low level of vibration in spite of the obvious vortex shedding from the jumper. The predicted vibration amplitudes in the cross flow vertical direction and cross flow horizontal direction are on the order of 1mm as shown in Figure 12. In order to characterize the motion of the entire jumper, Figure 13 plots the principal components (x,y,z) of standard deviation (STD) amplitude of vibration normalized by the jumper diameter. Note the STD amplitudes are less than 1% of the jumper diameter.

jumper and behind the buoyancy section. The vortex structures on the small diameter jumper itself are fairly well correlated, i.e. the structures along the length are at least somewhat in phase. However, the vortex structures behind the buoyancy section are not well correlated due to the presence of the strakes. Thus the buoyancy section, which might be expected to be the most significant cause of VIV acts more as a damper rather than a source of vibration. A cut though the buoyancy section showing velocity vectors is shown in Figure 15.

Figure 12 Jumper cross flow displacement at center

Figure 13 Jumper STD displacement at U=0.3m/s. Figure 11 Jumper in-line displacements at center Figure 14 shows flow results from Case 1 the jumper with contours of pressure on its surface and also pressure isosurfaces behind the jumper. These iso-surfaces surround vortex structures and thus are good indicators of vortex shedding. As shown, there are vortex structures along the entire length of the Following the initial characterization of the jumper response with a current speed of 0.3m/s we further tested the first case (0.3m/s) using 6 eigenmodes (rather than 12 eigenmodes) to characterize the structure. We also varied the time step in the simulation. After some preliminary runs we settled on a time step of 0.02s at a current speed of 0.3m/s and a structural description using the first 6 eigenmodes because we saw little

Copyright 2010 by ASME

diameter part of the jumper is 0.37Hz and the natural frequency of the 6th mode is 1.7Hz or 4.5 times higher. Because of the low vibration response in the first case we increased the current speed to 0.5 m/s for the second simulation. The change in current speed gave an increase in static in-line displacement to a maximum of 13cm near the jumper center. The vibration amplitudes also increased. However, the amplitudes are still small as shown in the plot of the STD displacements for this case (Figure 16). Note that the relative amplitude of vibration in the three principal directions has changed indicating that the relative contribution of the various modes has changed as well. Based on the predicted response for Case 1 and Case 2 we concluded that the jumper as configured with the buoyancy module and strakes will not experience large vibration amplitudes under the expected range of current conditions. The high density of the jumper and the action of the strakes tend to damp out vibrations that might be induced by vortex shedding from the small diameter pipe. Furthermore, the serpentine shape of the jumper and complex mode shapes mean that the vortex shedding forces are less likely to synchronize with the modal shapes in direction and phase. Finally, the large buoyancy section acts as a damper because it has a large associated added mass. We do believe however, that the vibration picture would be completely changed if the strakes were removed from the jumper. In this case, vortex shedding from the buoyancy section would force jumper vibration. However, lock-in is unlikely because the expected shedding frequencies would be below the natural frequencies of the structure.

Figure 14 Jumper with iso-surfaces of pressure showing vortex structure

Figure 15 Velocity vectors colored with velocity magnitude around buoyancy section difference from the response predicted when the higher eigenmodes were included in the structural model. Note that the expected primary vortex shedding frequency for the small

Figure 16 Jumper STD displacement at U=0.5m/s.

Cases 3 and 4: Jumper without buoyancy section (bare) with current speed of 0.5 m/s

Copyright 2010 by ASME

The next two cases examined a jumper without the buoyancy section and strakes. For these cases we changed only the geometry of the jumper but retained the structural characteristics of mode shape and stiffness. In Case 3 we chose a current speed of 0.5m/s or the highest current speed of interest. In Case 4, we chose a current speed of 0.454m/s in an effort to lock-in vibration response by aligning the vortex shedding frequency with the Mode 2 natural frequency. Case 3 - In the bare configuration, the jumper seemed to show a more correlated vortex shedding along its entire length than it did when the buoyancy section was present. Also, the bare jumper shows in increase in VIV over that seen when the buoyancy section as used. Overall normalized STD amplitudes increased to about 1% while it was about 0.4% at this current speed for the baseline geometry. This data is shown for a current speed of 0.5m/s in Figure 17.

Figure 18 STD Amplitude vs. distance along jumper at U = 0.454m/s (Case 4 - no buoyancy)

CONCLUSIONS AND RECOMMENTATIONS Two different jumper geometries were built modeling the baseline jumper design and a modified jumper in which the straked buoyancy section was omitted. In a series of tests with a constant current, neither design shows large VIV amplitudes. This is attributed to several factors. First, in the baseline design, the straked buoyancy section acts as a sea anchor and tends to damp out motions at the jumper center. The strakes are effective in preventing correlated vortex shedding from this buoyancy section so it does not contribute to VIV. In addition the vortex shedding frequencies from the small diameter sections of the jumper do not align well with the jumper natural frequencies. The jumper is heavy and stiff so that lock-in is limited to a narrow range of frequencies further limiting response. Finally, the serpentine shape of the jumper acts to reduce VIV. Those structural modes that might be excited involve motion of the entire jumper. However, only short sections of the jumper are aligned so that vortex shedding puts power into the system. For example, Mode 2 (Figure 2) has a primarily up and down response but only the horizontal sections of the jumper can act as power in regions. The vertical sections simply add mass to the system in this mode. These lessons carry over to the case of a bare jumper. We also found that a bare jumper is unlikely to experience large amplitude VIV. However VIV is increased significantly over the baseline case. Finally, it should be noted that we have only examined displacements in this study. A preliminary check of the simple bending stresses suggests that stresses are small. However, the small displacements may still result in significant stress in the jumper in some locations especially at the elbows in the jumper which have sharp corners.

Figure 17 STD Amplitude vs. distance along jumper at U = 0.5m/s (Case 3 - no buoyancy) Case 4 - Because the vibration amplitudes remained small we used this last case to explore the idea that the jumper was locking out when the vortex shedding frequencies did not align precisely with the natural frequencies of the jumper. There is some reason to suspect this because the bare jumper is relatively dense with a mass ratio of 3.2 It is well established that a high mass ratio limits the range of current speeds in which a freely vibrating cylinder locks-in and the vibration amplitude becomes large. We tested this idea here by choosing a current speed of 0.454 m/s. This gives a primary vortex shedding frequency of 0.54 Hz or the estimated natural frequency of Mode 2 in seawater. At this current speed the 2X harmonic of the shedding frequency is also close to the estimated natural frequency of Mode 4 in seawater. Case 4 does show some changes in the modal response as shown in Figure 18. The normalized STD amplitude in the flow direction increases from that of Case 3 from about 0.01 to about 0.015. We think the increase is noteworthy but this amplitude is not alarming.
2

Mass ratio is ratio of mass of an object to the mass of its displaced fluid

Copyright 2010 by ASME

REFERENCES 1. Spalart, P. R. and S. R. Allmaras, A one-equation model for aerodynamic flows, AIAA 92-0439 (1992) 2. P. R. Spalart, S. Deck, M. L. Shur, K. D. Squires, M. Kh. Squires, and A. Travin, A New Version of Detached Eddy Simulation, Resistant to Ambiguous Grid Densities, Submitted for publication in Theoretical and Computational Fluid Dynamics, 2005. 3. Holmes, S., Oakley O. H. and Y. Constantinides, Simulation of Riser VIV Using Fully Three Dimensional CFD Simulations, OMAE2006 92124 4. Constantinides, Y., Oakely, O. H. and Holmes, S., (2007), CFD High L/D Riser Modeling Study, OMAE2007-29151, Proceedings of the International Conference on Offshore Mechanics and Arctic Engineering.

Copyright 2010 by ASME

You might also like